Research work

Farruh Kushnazarov

19 August 2020

About Kushnazarov

Basic Information

Name


Contacts 		Kushnazarov Farruh
法如赫

173-2116-0407
k.farruh@bk.ru
Major Mathematical modeling, numerical calculation and coordination of computer programs
Highest Education Ph.D. from St.Petersburg National Alexander I Emperor Transportation University
Languages English Russian Chinese
Excellent Native speaker HSK2 (elementary)

Bachelor and Master

Kushnazarov Farruh’s favorite subject from freshman year was mathematics and communication network system. Because of this, his final theses for bachelor and master degrees were about the communication network system. After successfully achieved a master’s degree with a high score, he got an offer from the university to continue researching the communication network. Feeling the need for practical experiences, he started working as an assistant professor and programmer for two years, during which he collected remarkable experiences and data for further research.

PhD

Finally, in 2012 he back to university to apply Ph.D. and work with a well-acknowledged professor in communication network and virtualization fields. As a researcher, he not only considered in the theoretical part but worked with several companies for applications as well. Moreover, each result was provided for implementations on data-centers, smart-city, and other projects. In Russia, especially Saint-Petersburg researcher Kushnazarov Farruh’s papers were frequently cited and discussed in dozens of conferences by other researchers. Kushnazarov Farruh has given 144 hours of lectures on communication network system for students and published two textbooks as the first author.

Nowadays

  • In 2016 he moved to Shanghai for working as a data scientist in a well-known online education company. He is working on different machine learning and deep learning projects—his dozens of projects brought enormous benefits to the company.
  • Even with heavy-loaded work, every year, he publishes scientific articles on artificial intelligence and communication network fields in international conferences, including the global academic platforms such as IEEE, ACM, and others.

Goal and Mission of DLL Protocols

Research Goal

Based on the variation of the basic parameters of the protocols and transmission procedures to develop and evaluate the performance of communication protocols that provide guaranteed data delivered over communication channels with noise.

Research Missions

  • Analyze the FEC (Forward Error Correction) for detecting errors and recovering information in communication channels with noise, taking into account the determination of the main factors.
  • Define the impact of noise on network performance on the different ARQ (Automatic Repeat Request) modes.
  • Develop a mathematical model of the proposed method for evaluating performance, taking into account the peculiarities of computational algorithms for implementing FEC for different ARQ modes to increase Shannon capacity.
  • Develop recommendations for data link layer protocol’s settings to achieve a given level of performance.

Scientific Novelty

Encoder and Decoder

Implemented Convolutional Encoder and Decoder

To experimentally determine the Viterbi decoder’s operability with the offered method using the Matlab environment (Simulink, Communication Toolbox package), a communication channel model was built. The parameters of the Viterbi decoder are taken as follows: R = 1/2, d = 7, L = 19. The input data stream in the form of a sequence of pseudorandom numbers, the frame size is 1120 bytes.

Fig.1 - Simulation Viterbi Decoder on Mathlab: Communication Toolbox

Result of the Research

A methodology has been developed for calculating the real data transfer rate for the ARQ methods Stop-and-Wait, Go-Back-N, Selective Repeat based on the frame sizes, the number of frames in the window and the level of noise in the communication channels. Recommendations are given on setting up a timer for various data transfer methods.

A simulation computer modeling of data transfer processes using CRC codes in two environments (C/C ++ and Matlab) was carried out under conditions of varying values of the noise level in the data transmission channel. The necessity of choosing the type of standardized generating polynomial, which can reduce the probability of failure to detect errors in frames, is substantiated.

Simulated data transfer process using a convolutional coder/decoder to determine the probability of the incorrect decoding of received data is carried out. The expediency of using the technique of interleaving symbols to reduce the level of this parameter is proven.

The results of simulations and implement models on the EMC2 data-center, and SmartCity projects of Moscow are compared. It is shown that their deviation does not exceed the boundaries of 11%, which confirms the consistency of the research.

Implementation of Research Results and Major Publications

Implementation of Research Results

  • Implemented at the company L-Cube LLC to optimize the performance of the departmental network;

  • Performed at the company Domus Sapiens LLC to improve the quality of data transfer in a wireless network;

  • The tested result on the EMC2 data-center.

EMC2 data-center
Fig.2 - Implementation Certificates

Major Publications: During 2013 – 2020:

Fig.3 - Patents

Invitations for Further Researches

Fig.4 - Recomendation Letters

Future Plan

The main expected results of the research can be summarized as follows:

  1. Analysis of existing methods for increasing the throughput protocols showed their limitations in taking into account the noise in the communication channel and not implementing AI to reach the required throughput.
  2. To define significant factors, which impact the throughput of the cellular network.
  3. The method decreases the latency by incorporating AI to HARQ for different modes, based on the historical data.